Map multiple points within distance of multiple sites

[Drewz]

Hi All,

I have a table of crime stats and a table of CCTV sites, see example tables below. I'd like to map crimes around each CCTV site, for example within 500 m of each site as shown below (although not with the site labels, that was just for clarification), 

I can apply the haversine calculation (as shown be Phil Seamark in image below). I assume the best option is to add a calculated column to the crime table. I tried below, but I get a range of errors. I'll also need to add criteria to select the closest CCTV site in case there are two with 500 m, but thought I'd focus on the first part. 

Any thoughts on my approach or my calculated column?

CCTV Sites =

IF(

CALCULATE(

12742000 * ASIN((SQRT(0.5 - COS((CCTV(Latitude)) * DIVIDE( PI(), 180 ))/2 + COS(CCTV(Latitude) * DIVIDE( PI(), 180 )) * COS(Crime[Latitude] * DIVIDE( PI(), 180 )) * (1-COS((Crime[Longitude]- CCTV(Longitude) * DIVIDE( PI(), 180 )))/2)))  < 500), CCTV (ID),"")

CCTV Installaton
Camera ID	Long	Lat	Installation Year
201	-36.9722	174.7867	2019
202	-36.9651	174.795	2017
203	-36.9665	174.8085	2018
204	-36.9519	174.8385	2020
205	-36.9656	174.8156	2019

Crime Stats
ID	Long	Lat	Crime Type	Year	Calculated Column
1	174.7902527	-36.97949982	Property	2015
2	174.7902527	-36.97949982	Injury	2021	203
3	174.8324585	-36.96062088	Cyber	2015
4	174.8324585	-36.96062088	Injury	2017
5	174.8353424	-36.93717957	Injury	2020
6	174.8353424	-36.93717957	Injury	2016	201
7	174.8353424	-36.93717957	Cyber	2020
8	174.8505554	-36.9417305	Cyber	2019
9	174.8505554	-36.9417305	Cyber	2018
10	174.8505554	-36.9417305	Property	2018
11	174.7889252	-37.00097275	Injury	2020	205
12	174.7889252	-37.00097275	Property	2016
13	174.7889252	-37.00097275	Property	2018
14	174.8352814	-36.93664932	Injury	2019
15	174.7895508	-36.97363281	Injury	2018	202
16	174.7895508	-36.97363281	Property	2016	202
17	174.7895508	-36.97363281	Property	2017
18	174.8317566	-36.96046066	Injury	2022
19	174.8317566	-36.96046066	Cyber	2021
20	174.8399353	-36.94024658	Injury	2017

[lbendlin]

see message 6

[Drewz]

Ahh, yes I did miss message 6. With large data sets, I'm getting some unusual banding when I specify distance of 200 m (see image below). Also, the haversine formula that you used gives quite different distance results. At least a factor of 2 different. I thought they would be similar. I checked the original formula by measuring a distance of Google Maps, and it gave me a matching result within 1 decimal place.  

[lbendlin]

I refactored the formula to its more common format - see if that is better. I also included a simpler version that gives the same result.

DC2 = 
var a = summarize(CCTV,CCTV[Camera ID],CCTV[Lat],CCTV[Long],"Dist",12742000 * asin(sqrt(power(sin(RADIANS(CCTV[Lat] - Crimes[Lat]) / 2), 2) +  power(sin(RADIANS(CCTV[Long] - Crimes[Long]) / 2), 2) *  cos(RADIANS(CCTV[Lat])) * cos(RADIANS(Crimes[Lat])))))
var b = TOPN(1,a,[Dist],ASC)
return CONCATENATEX(b,[Dist])

DC3 = 
var a = summarize(CCTV,CCTV[Camera ID],CCTV[Lat],CCTV[Long],"Dist",6371000 * acos(sin(RADIANS(CCTV[Lat]))*sin(RADIANS(Crimes[Lat]))+cos(RADIANS(CCTV[Lat]))*cos(RADIANS(Crimes[Lat]))*cos(RADIANS(CCTV[Long]-Crimes[Long]))))
var b = TOPN(1,a,[Dist],ASC)
return CONCATENATEX(b,[Dist])

Your banding is likely caused by picking a column from the wrong table.

[Drewz]

Sorry for the delay responding. I've been testing the above. I was getting some spurious results, so I stuck with the original haversine formula. That combined with your "brut force it" method and a distance slicer worked. So, thanks very much for your help.

Ideally I would like to simplify the code so I only have one calculted column that returns the CCTV ID when less than 500 m away. So if anyone has any suggestions to refine the code it would be greatfully received. Otherwise I'll keep trying to figure out some options.

[lbendlin]

 that returns the CCTV ID when less than 500 m away. 

You can only decide that after you have computed the distance.So there is no advantage in throwing away the result. Imagine you then want to say "less than 1000m" etc.  Keep the actual values.

One thing you can consider is to ditch the Haversine and do the simple flat earth computation based on the average meters per degree for your latitude and longitude area.  Much easier to compute.

[lbendlin]

1. what are the errors?

2. wouldn't you want this as a measure, rather than a calculated column?

[Drewz]

Hi Ibendin,

It's not recognising the references for longitude / latitude for each table. Even for the Crime[Latitude] and Crime[Longitude] is seems to be wanting a measure. I put a screen shot below. I think it's something to do with the fact that the tables are unrelated and I need to do a row by row operation. 

I would have thought a calculated column would be better for a row by row operation. Also, it can process the information during the refresh which means it should respond quicker (my basic understanding). But obviously my DAX isn't right. 

Basically I just want to add a CCTV site in the calculated column if it's within a certain radius of the crime location. In my head it makes sense, but I'm not sure how to get it to work with unrelated tables.  

[lbendlin]

You can combine both tables. Add a Location Type identifier to distinguish between cameras and crimes.

Combined = UNION(SELECTCOLUMNS(CCTV,"Lat",[Lat],"Lon",[Long],"Location Type","Camera","Crime Type",BLANK(),"ID",[Camera ID]),SELECTCOLUMNS(Crimes,"Lat",[Lat],"Lon",[Long],"Location Type","Crime","Crime Type",[Crime Type],"ID",[ID]))

I also see a very interesting error message - Your CCTV sample table has the wrong column headers.

 see attached

[Drewz]

Hi Ibendin,

thanks for the example. I'm less worried about displaying the CCTV location and more interested in selecting the crimes within a certain radius based on the haversine calculation. Eventually I want to campare crime rates within a radius of CCTV locations vs crime rates beyond that radius to see if there is a difference. I think to do that, I'll need to have a calculated column in the crime table that adds a CCTV site if its within a radius. Any thoughts. 

Oh, and thanks re the headers.

Cheers Andrew

[lbendlin]

I would brute force it , calculate the distance from a crime location to all camera locations and then pick the closest.  Storing that distance will probably be useful too.

see attached for a possible implementation.  Note that DAX has a RADIANS function.

Note that the Haversine distance is rather meaningless in an urban environment.

[Drewz]

Lol @ "brute force it". I'm open to anything. although for rural locations the nearest CCTV location may be miles away. But I could then filter them out. So how would "brute force it" look? Would it be a calculated column added to the Crime table?